I’m not an anti AI person, is very useful. But I can’t see any way to stop AI from destroying the world. For who doesn’t believe here are some short examples:
AI blackmails 1-n scientist in a lab to do X.
AI pays a guy/mercenary/drug-addict to kidnap the family of 1-n scientists to do X. (If you say impossible, AI can promise some money upfront plus money later, eventually someone will do it, might not be perfect, but in 100 attempts one will work, even if the AI gets scanned 99 times)
AI uses 0 days to hack and access confidential info to blackmail or pretend to be someone else and give some instructions.
X can be: release deadly toxin that can kill people.
All this to explain the threat is real, in case you’re dismissing it. the point of my question is how to slow it down.
Ideas so far:
Disrupt T in a way that can’t produce semiconductors: will stale production for a few years. Very hard to rebuild.
Disrupt K and J in a way that can’t peoduce HBM: without the high speed memory T can’t create AI chips. medium difficulty to re build.
what other ideas do you have? what choke points could disrupt AI progress the most?
i don’t believe in political ideas such as “let’s not build more datacenters” or “let’s all agree t o don’t train more advanced AIs”
Variables:
T: Taiwan
K: Korea
J: Japan
[link] [comments]